Santiago A. Utsumi, PhD
Innovations for Sustainable
Farming & Ranching
Inspired by Nature, Driven by Science
​The rise of "Agricultural Technoecology” as a new field for agribusiness and science
Grasslands represent the largest percent of land entrusted to mankind for provision of agricultural services. yet they are exposed increasingly to growing global stress. As the use of grasslands shift or intensify, the provisioning of services -for example, buffering climate change effects, or provision of healthy food systems- will rest increasingly on our capacity to bring up-to-date solutions for agribusiness, food supply chains, and global trade.
In response to this pressing demands I believe we have nurtured, almost unintentionally, a new field for interdisciplinary sciences, the technoecology of agricultural systems. Broadly, I define this field by the application of emergent sets of technologies (instruments, products and processes) that allow answering fundamental questions about living organisms and the environment in ways we never could to answer y means of traditional field observation alone. Here I share some examples to contribute to the foundation of the discipline.
Example 1: Dairy cattle links optimal assessment and environmental trade offs to regulate feed intake
Perhaps the best example I have comes from a "plant-pot and foraging" experiment I've conducted back in 2001, and years later scaled by my students in a robotic cow management study. Here are some highlights of that work. The decision making of dairy cows regarding foraging (when and where to eat, and for how long) can be affected profoundly by offering them arrays of feeds that differ spatially, temporally and nutritionally. I tested this hypothesis by offering dairy cows arrays of fescue and alfalfa patches. Plant species, of course, differed in their preference by cattle (cows liked the alfalfa over fescue) and potential for intake rate (bites were larger and heavier and intake rate faster on alfalfa than fescue, at low depletion rates). I controlled experimentally the "selection effort" through manipulation of the forage patch size and spacing among forage patches. Can you guess what might happen when I challenged cows to forage in arenas that contained different patch size and spacing arrays of alfalfa and fescue?
Before I go any further, I want to be sure that you are convinced about your answer, otherwise I invite you to watch the videos again. Here is the hypothesis that I was answering, Cow traffic will change profoundly, as feeding rates change, spatially (how far cows must walk for feed) and temporally (how often cows reach fresh feed); both manipulations will matter almost equally as ruled out by the physics of movements deeply rooted on Albert Einstein's theory of relativity: "for a subject that moves in a time-space domain, the faster the subject (cow) moves through time, the shorter that subject will move through space (farm), and vice versa", Going back to the experiment and hypothesis, the longer the distance a cow walks in space, or the longer the time interval this cow experiences between feeding events, the longer she will stay eating using longer meals, thereby reducing almost instantaneously the opportunity for being too choose or selective. Thus nutrient "optimization" (i.e.solving constraints to acquire a balanced intake of nutrients) drives changes in behavior profoundly, with implications to feeding management.
Imagine now that you are given access to real-time devices to track robotic dairy cows, spatially and over time. The following would be a very close picture of what you will see. As time and space domains shift and the feeding context and experience differs, cow traffic will change accordingly. Furthermore, feeding rates, cow traffic and robotic milking will vary markedly as our feeding management routines change; yet, remember that a cow's time and space perspective are indivisibly connected; It would be expected cows to increase meal frequency (either of grass or PMR), reduce meal size (i.e. feeding duration), and increase robotic milking frequency if they are given access to feed more frequently, both spatially by walking closer to feed resources, or temporally by accessing feed resources sooner. This modification of feeding behavior and milk performance could be achieved by frequently offering fresh grass breaks (so called multiple-way, or A-B to A-B-C-Z... systems), fresh PMR drops, or more frequent top-dressed PMR pushes (i.e. use of robot mixers and feed pushers) throughout the day, and "at close distances", otherwise cows will trade-off challenging feeding efforts (i.e. layout restrictions) against the benefit of reaching such feed, and even when those feed options are nutritionally complete and preferred.
Interestingly this is what my student Kate Steensma discovered in the spring and summer of 2011 when she was provided with the challenge to scale-up the plant pot experiment to field and farm scales. In this experiment robot dairy cows clearly preferred a 7-species mix over a 2-species mix when both feed treatments were offered adjacently. Yet, selection for the 7-species disappeared and cow traffic was significantly reduced as daily feeding sequences containing the two treatments were space further apart and away from robots. Thus, walking effort and frequency of reaching fresh feed outweighed any likely benefit of being choose or selective, on robotic cow traffic and milk production. The map below gives a broad and fair approximation of what Kate Steensma saw for her MS thesis on the robotic cow traffic and milk production experiment.
As corollary for this integration between the ecological understanding of herbivory and the application of GIS to address fundamental questions in agriculture, I leave you with the following reflection. We have upfront a rich horizon for synthesis work to integrate research-based knowledge and farm observations for deployment of monitoring and management gadgets in ways that they were not possible before. This integration opens new opportunities to revisit and strengthen existing theory and to use such theory in the format of mathematical models, either to predict and test impacts, or to serve as bases for new Smart Farming Platforms (SFP) and decision-making tools. Following the seminal work by Stephens and Krebs back in 1986, I drawed below a conceptual framework as model application to explain how robotic cow traffic and production might improve, assuming that main physical constraints and external factors (robot layout and barn design) can be equally controlled (see Robotics page).
Example 2: Robotic milking and cow psychology, personality and production
Skinner's rules for conditioning learning are now robotic. One question most dairy farmers that transition from conventional milking to robotics usually ask is, what's the best method to introduce cows to a robot? As I reflect over and over on the subject I usually advise with findings that are deeply rooted on experiments and readings on animal behavior I have conducted while I was a Ph.D. student at NMSU. I remember my advisor, Dr. Andres Cibils challenging me from the go with serious Skinnean literature of psychology and behavior. This readings helped me profoundly to deepen my understanding about how we can use prescribed practices to train herbivores to eat some foods or avoid given habitats, and regardless of whether foods might be highly preferred, or habitats were previously indexed as familiar, comfortable and safe. Here are few discoveries on the subject and how we can apply them to robotic cow traffic and milk production management. Going back to the farmers question then, how a cow “feels” upon first robotic milking experiences rules out much of the success of that individual as future robot expert. This behavior rule has a scientifically proven solid base on the quadrants of “conditioning learning”. Assuming your robot layout, and pasture or barn design are well suited for the task (either new or retrofitted facilities), the main drive for robot cow traffic would be “appetite" (not hunger!) and the rewarding desire to eat arrays of feeds that truly "nourish and satiate". Furthermore, if we apply Dr. Skinner's teaching to manage robotic cows, results will be as follows:
​
1) adding high quality robotic feed with rich nutrient-flavor associations (cap of 10 kg/cow/day of robot pellet feed) will increase robot visitations through a conditioning phase known as positive reinforcement. Conversely, by reducing the amounts, quality or both characteristics of the same robotic feed pellet we will reduce the value of a cow's robotic milking experience; thereby reducing the frequency for future robot visitations. This conditioning phase is what behavior experts define as penalization or negative punishment.
​
​
2) a sequence of poor robotic milkings experiences, for example by forcing physically cows to enter the box by fear, will act as a true punishment action and will certainly reduce a cow's attitude for robotic milking. This acquired condition is called positive punishment. Conversely, relief from such fear through use of low stress handling practices has potential to revert low robotic milking visitations, but certainly this mending action will be less advantageous compared to using positive reinforcement through feeding from the go. This relief conditioning phase is called negative reinforcement.
​
As a general advise and gold rule, farmers should consider minimizing cow contacts and situations of animal fear with regard to robots. Enticing cows to visit robots frequently through positive reinforcement is the best scenario that farmers would have to train and manage high performing robotic dairy cows.
